The Buy Me Once Longevity Awards
How We Judge: The Weakest Link Method
Every award we give can be checked, challenged and scrutinised. This page is the complete published methodology - the same document our own researchers score against - and the five checks every category must pass before we announce a single winner.
The principle
Every product category has predictable failure points - the reasons people throw products away and buy replacements. We call these Weakest Links. A kettle's element burns out. A frying pan's coating flakes. A jacket's zip fails.
The Longevity Awards don't ask "which product is best?" They ask: which product has done the most to eliminate the reasons people throw this product away? That shifts judging from taste to structural analysis: what breaks, how often, how badly - and what the manufacturer has actually done about it.
Then we try to prove ourselves wrong. Research proposes a winner; it doesn't crown one. Every category must pass five proof checks, built so that whoever did the research never gets to mark its own homework.
The process
Step 1 - Identify the Weakest Links
For each category we conduct a forensic analysis of how products really fail, drawing on reliability surveys (such as Which?), the Open Repair Alliance's open dataset of more than 400,000 repair records from repair cafés and community repair events, repair-trade insight from people who fix these products for a living, warranty and complaint data, teardowns, spare-parts listings and long-run owner reports. Each failure point is documented with what fails, why it fails, how often (Frequency, 1–10) and how badly (Severity, 1–10 - does it kill the product or merely degrade it?).
Not all evidence counts equally. When sources disagree, this is the order that wins, and each published scorecard states which level each score rests on:
| Rank | Evidence |
|---|---|
| 1 | Independent long-run reliability datasets (large owner surveys, failure statistics). |
| 2 | Verified warranty, repair and complaint records - including large open repair datasets. |
| 3 | Teardown and materials evidence - what the part is actually made of and how it's built. |
| 4 | Repair-trade testimony from people who fix these products for a living. |
| 5 | Aggregated owner reports and repair-community records. |
| 6 | Manufacturer claims - accepted only with documentation (part numbers, published policies, spares listings), and weighted lowest. A slick spec sheet earns nothing on its own. |
We also correct for where evidence comes from: repairers see the products that break, not the ones that don't, and enthusiast communities over-represent cult brands. Anecdote is never treated as frequency data - frequency scores rest on levels 1 and 2 wherever they exist. Every row of every hierarchy must be backed by at least one independent source (we aim for three), and we publish which.
Step 2 - Rank by importance
Each failure point's Importance = Frequency × Severity. Ranked, these produce the category's Weakest Link Hierarchy - a prioritised list of the failures that most determine real-world lifespan, published alongside every award.
Step 3 - Score the engineering
Every candidate is scored on how well it defeats each failure in the hierarchy, on one scale. Two principles are built in:
Prevention beats repair. A part that never fails beats a part you can fix - because even the best repair forces a fix-or-replace decision a durable part never does.
Repair is discounted by faff. A repair that needs an engineer, real money, or a fortnight without the product is worth far less than the marketing implies - because that's the point where most people give up and rebuy.
| Score | What earns it |
|---|---|
| +5 | Failure materially designed out - not reasonably expected to occur under typical use within the category's stated lifespan benchmark. |
| +4 | Highly durable by material or design - lasts many times longer than the category baseline. Or a part the user renews themselves, free and genuinely faff-free. |
| +3 | Notably more durable than the baseline. Or user-repairable with parts and clear instructions. |
| +2 | Moderately more durable. Or repairable but with real friction - agent repair, meaningful cost, sourcing hassle. |
| +1 | Marginally better built. Or warranty-replacement / high-faff repair only. |
| 0 | The category baseline. Every score is relative to this. |
| −1 / −2 | Worse than the baseline, or introduces a new failure point. |
Step 4 - Calculate the score
A solution to the most important failure counts for more than a solution to a trivial one. Because the maximum varies by category, scores are also expressed as a percentage of the maximum achievable - and you'll notice even winners score far below 100. That's deliberate: 100 would mean every failure designed out entirely. The gap between the winner and 100 is our published brief to the industry.
These are structured judgements, not laboratory measurements - so we stress-test them. Before any verdict is locked we run a sensitivity check on every figure, not just the disputed ones: each Frequency and Severity is moved up and down by one and two points, each product score by one point, and the category is re-ranked every time. Any single figure that could change the winner, the Highly Commended list or a joint verdict is flagged as decision-bearing, and it must then be backed by three independent sources, at least one of them a working repairer, before the verdict can stand (see check 4 below). If the verdict doesn't hold on the repairers' own numbers, it doesn't hold.
Step 5 - Select the winners
The highest-scoring product on the published longlist wins. Two products are declared joint winners only when both parts of a stated rule are met: the gap between their totals is no more than 5 percentage points of the category maximum score, and the sensitivity check cannot separate them. Genuinely excellent runners-up are named Highly Commended. Where scores are close but not joint, the tie-break is a stated rule: first, the product that best solves the highest-ranked weakest link; second, the product whose longevity solutions are most accessible to everyday consumers.
Step 6 - Prove it: five checks before anything is announced
A proposed winner is not an award. No category is announced, pitched to press or offered a badge until it has passed all five checks below, and the person who wrote the research never grades it. Nothing passes on someone's say-so: every check closes only with a receipt - the source, the date, and what it showed.
The five checks
- Did we follow our own method? Each category's research is checked, item by item, against this page: the category definition and baseline, the hierarchy and its evidence, model numbers pinned, the full grid, exclusions with reasons, sources, conflicts of interest disclosed, and the date the brand accuracy check was sent and closed.
- Is every fact true today? Every factual claim carries a receipt: the page it came from, the exact wording relied on and the date it was checked. An automated check re-opens every source and confirms the wording is still there, falling back to the public web archive where a page has moved. Then a separate AI fact-checker, with live web search and no stake in the research, tries to disprove every claim. A dispute is never settled by picking the answer we prefer: we go back to the primary source - the manufacturer's spares page, the survey, the manual - and record what it says.
- Did we miss anyone? An independent sweep lists every brand on sale new in the UK from four angles: mainstream retail; small, independent, heritage and B Corp makers; premium imports through UK specialists; and the brands the repair trade rates. Every brand it finds is either scored into the grid or given a written reason for leaving it out. Only mass-market products priced at or below the category baseline may share one reason; every independent, premium or repair-trade brand gets its own.
- Can every number be justified? Every row of the hierarchy is backed by independent evidence (repair-café records, reliability surveys, spare-parts and service listings, regulatory data, repairers). Every decision-bearing figure from the sensitivity check needs three independent sources, including at least one human repairer, and every hierarchy row and podium score is challenged against repair-trade and manufacturer evidence. Then we ask one blunt question: if every challenge were accepted, would the verdict survive?
- Is it safe to launch? Checks 1 to 4 pass, the winner's availability new in the UK has been verified within the last seven days, and the brand accuracy-check window has closed. Only then do we announce the winner.
It's already earned its keep. On kettles alone, these checks found a model we had excluded as discontinued back on sale, a repair centre described as authorised that turned out to be independent, a stocked £3.60 spare lid that raised one product's score, and repair-café data showing limescale is a far smaller killer of kettles than the grid had assumed. Each finding was fixed, with a receipt, before a word went public.
Every award is held to this standard, including those already published. Any change these checks force is logged publicly as a correction, with the date and the score impact.
After the award: we keep checking
A verdict is locked at announcement, but the facts under it are not left to go stale. Every week, each winner and commended product is re-checked for availability new in the UK, price, recalls and administration news, and every cited source is re-opened to catch broken links and changed wording. Every month, the facts behind every announced award are challenged again with live web search. If a winning model is discontinued, we look first for the same brand's direct replacement and assess it against the same scorecard before considering any other product. Anything that changes a published finding goes through the public corrections process below.
Who scores, and on what evidence
The awards are built on real-world failure-evidence synthesis, not destructive lab testing. We don't load washing machines with bricks until they die; short lab tests can't reveal multi-year durability anyway. We weigh the best available evidence of how products actually fail over years of real use - and the people who see that most clearly are the ones repairing these products day in, day out. Where hard data exists, data leads (see the evidence hierarchy above). Where judgement is required, the judgement is shown: every score publishes with its evidence level, sources and reasoning, so you never have to take a number on trust.
Who scored the 2026 series: the 2026 verdicts were researched and scored by the Buy Me Once research team, led by founder Tara Button, against this published method. We say that plainly rather than hiding behind an unnamed "expert panel" - and it's why the full working publishes with every award: a scorecard anyone can check is a stronger claim than a title nobody can. Every 2026 category is also reviewed by at least one working repairer before it's announced: repairers check our failure grid from the bench, tell us which numbers they'd change and what we've missed, and the verdict has to hold on their numbers. Repairers who help us can be credited, with their permission, in our Recommended Repairers directory. From the 2027 series, each category adds an independent assessor panel drawn first from the repair trade, with the number of assessors, their roles and credentials, recusals and disagreements published per category. An assessor with any commercial or personal connection to a candidate brand is recused from all comparative decisions involving that brand - and from the whole category where the conflict is material; no assessor is paid by, or answerable to, any candidate brand. Repair a category for a living? We're recruiting.
How we obtain products: honestly, and on a small company's budget. Assessments are built on the documented evidence above - design specifications, spares listings, manufacturers' written answers to our technical questions, and the independent failure record. We obtain samples where we believe a product may be the best in its category; when a product has been physically examined, the scorecard says so and states whether it was bought at retail or supplied by the brand. Brand-supplied samples are always declared, and a sample never substitutes for independent failure evidence.
Eligibility gate: a product must be currently on sale new in the UK to hold a current-year award - we re-verify availability before publishing and every week after, because an award for a product you can't buy helps nobody.
Eligibility & how candidates are found
Any product on general sale new in UK retail is eligible. There is no fee to be considered, and no product is excluded for lacking a relationship with Buy Me Once. Our research hunts where the best-in-category is most likely to be found - failure data, repair-community reputation, independent testing - and the independent market sweep in check 3 makes sure small and independent makers aren't missed. Even so, any brand can ask to be judged: if you believe you out-engineer a current winner, request assessment and we'll score you against the published hierarchy on identical terms. Winning is never contingent on asking, paying, or being sold by us.
The longlist is published. Every category verdict lists every product screened, the date of the market search, and why any product was excluded - so "maybe they never even looked at X" is a question you can answer yourself, from our own records.
Right of reply & corrections
We'd rather be corrected than wrong. Before publication, every manufacturer whose product is named in a comparative assessment - winner, commended, or criticised - is contacted for an accuracy check on the technical facts, with at least 14 days to respond. It is a check on the facts, not approval of the verdict: new, verifiable facts from a brand (a guarantee we hadn't found, a spare part we missed) can change a score, but no brand can veto a conclusion, and a non-reply doesn't block publication. After publication, anyone - brand, repairer, or reader - can challenge a finding through our standing corrections channel. We review challenges against the same public criteria, correct the record when we're wrong, and log every correction publicly: the date, the issue, the decision, and any score impact. This is free, always. No brand can pay to change, raise, or remove a score.
Annual cycle, permanence & when an award can be withdrawn
The awards run annually and every award is year-stamped. A 2026 winner is a 2026 winner forever - a later, better product doesn't retroactively strip it, just as a Which? Best Buy isn't revoked when next year's model beats it. In later years other products may win; that's the standard improving, not the past being erased. New categories are added through the year as each one passes all five checks.
Awards attach to the exact model, not the brand. Each award names the specific model number assessed. If a product is materially changed after its award - a different element, coating, compressor, factory or key supplier - the award does not automatically transfer to the new version, and licensed winners are required to notify us of material design changes.
Permanence has honest exceptions. We will suspend or withdraw an award if the assessment is shown to be materially wrong, if a safety issue emerges, if an undisclosed conflict comes to light, if the awarded model is materially changed without notification, or if the badge is used in a misleading way. Historical awards stand unless materially flawed or misused - and any withdrawal is published, with reasons.
Independence & where the money goes
The award is worthless if it isn't trusted - so rather than asking you to take independence on faith, here is exactly how the money and the judging are kept apart:
Where the money goes
We earn money from some winning products. Buy Me Once is a shop: we sell some winners directly, and some product links on our pages are affiliate links that pay us a commission. Every award page discloses, per product, whether we sell it or earn commission on it, and each category's conflicts are recorded as part of check 1 before it can launch. We don't buy stock speculatively - what we list follows what the research finds, and we delist products that stop meeting the standard - but we don't pretend the income doesn't exist. The safeguards below are how we stop it mattering.
Scores lock at announcement. Each category's verdict is locked at the moment it is announced: from that point the year's scores are final, and can change only through the public corrections process - never through a commercial conversation. No award-linked commercial discussion of any kind - badge licensing, a new listing, a commercial term - happens with any brand before its category's scores are locked. Where we already sold a product before its category was judged, that pre-existing shop relationship is disclosed on the award page. Each scorecard publishes its lock date.
The badge licence is display rights, never entry. After - and only after - a product has won, its brand may license the Winner mark for its own packaging and marketing, at a published fee schedule banded by company size - a small independent pays less than a multinational, and no fee is ever negotiated per result. Declining changes nothing: the award, the listing and our promotion are free and permanent either way, and a licensed winner can still be unseated by next year's research. Licensing is handled after verdicts are locked and has no route back into scoring.
We regularly commend products from brands we have no commercial relationship with, because the engineering deserves it - and we say so on every award page.
What the awards do NOT judge
Aesthetics - a product can be ugly and win. Purchase price - expensive products get no extra credit; cheap ones no penalty. (Repair cost is different: it is scored, because a £90 repair on a £100 product is the moment it gets binned - that's a longevity fact, not a price judgement.) Sustainability beyond longevity - we don't audit supply chains or carbon footprints; other certifications do that well. Our lens is narrow and deep: will this product last? Brand reputation - a household name gets no advantage; a startup no disadvantage.